This invention relates to a transmission system having a multiple-input multiple-output (MIMO) architecture. More particularly, the invention relates to a multiple-input multiple-output transmission system, which has a plurality of transmit antennas and receive antennas and is capable of high-speed data transmission, and to a transmit station and receive station in this system.
Of special interest in present-day wireless communication systems is a spatial multiplex transmission technique in which transmission capacity is increased in proportion to the number of transmit antennas by transmitting different data streams in parallel from different transmit antennas. Different transmit antennas are arranged so as to be uncorrelated with one another and the data streams transmitted from each of the antennas are received by receive antennas upon traversing independent fading propagation paths.
By utilizing a plurality of receive antennas arranged so as to be uncorrelated with one another to establish a multiple-input multiple-output (MIMO) system, a channel correlation matrix having a high degree of freedom can be generated and it is possible to improve SNR when the spatially multiplexed data streams are demultiplexed.
FIG. 14 illustrates the architecture of a MIMO system, in which TR represents a transmit station and RV a receive station. Data streams S1 to SM the number of which is the same as the number M of transmit antennas ATT1 to ATTM are transmitted from the transmit antennas ATT1 to ATTM upon undergoing processing such as data modulation, oversampling, D/A conversion, orthogonal modulation, frequency up-conversion and band-limiting filtering in respective ones of transmitters TX1 to TXM. The signals transmitted from the antennas ATT1 to ATTM pass through independent fading channels hmn (m=1 to M, n=1 to N) and are multiplexed in space, after which they are received by N-number of receive antennas ATR1 to ATRN. The signals received by the receive antennas undergo processing such as filtering, frequency down-conversion, orthogonal detection and A/D conversion in receivers RX1 to RXN, whereby receive data streams x1 to xn are generated. Since the receive data streams are in a form in which M-number of transmit data streams have been multiplexed, signal processing is applied to all of the receive data streams to thereby demultiplex and reproduce all of the transmit data streams.
FIGS. 15 and 16 illustrate examples of the structures of a transmitter and receiver in a conventional digital wireless communication system that does not include a MIMO system. The zone of the transmitters (TX1 to TXM) and the zone of receivers (RX1 to RXN) shown in FIG. 14 are enclosed as TX and RX by dashed lines in FIGS. 15 and 16, respectively. In stages ahead of the transmitter TX (FIG. 15), the transmit data is encoded in accordance with a prescribed encoding scheme, and the encoded data is mapped to two orthogonal axes I, Q in accordance with the modulation scheme (QPSK, 16QAM, 64QAM). Next, the transmit data has its pilot time-multiplexed and is then transmitted from a transmit antenna ATT upon being subjected to oversampling, D/A conversion, orthogonal modulation, frequency up-conversion and band-limiting filtering.
Filtering, frequency down-conversion, orthogonal detection and A/D conversion processing are executed in the receiver RX (FIG. 16). This is followed by channel estimation, synchronous detection, data demodulation (mapping) and data decoding processing.
Algorithms of signal processing of a data processing unit DPU, which demodulates the transmit data streams S1 to SM (FIG. 14) from the receive signals, include a linear algorithm referred to as ZF (Zero-Forcing) or MMSE employing a matrix that is the inverse of a channel correlation matrix, and a non-linear algorithm typified by BLAST (Bell Laboratories Layered Space-Time). Also known is a method such as MLD (Maximum Likelihood Decoding), which does not use a matrix that is the inverse of a correlation matrix.
The following relations hold if the transmit data stream is represented by an M-dimension complex matrix S and the receive data stream by an N-dimension complex matrix X:X=HS+V E[VV* ]=σVI where E represents an ensemble average, H an N×M complex channel matrix (h11 to hMN), and V a complex white-noise matrix of average value 0 at a variance σV. The “*” symbol represents a complex conjugate transposition of a matrix. Further, I represents an N-dimension unit matrix.
With the ZF algorithm, a transmit data stream is estimated according to the following equation:Ŝ=(H*H)−1H*X where H*H is referred to as a “channel correlation matrix”. Since a matrix that is the inverse of the channel correlation matrix exists, the relation N≧M becomes necessary.
With the MMSE algorithm, a transmit data stream is estimated according to the following equations:Ŝ=(H*H+αI)−1H*X α=σV/σS=M/ρE[SS*]=σSI where ρ corresponds to the SNR per receive antenna. With MMSE, it becomes necessary to estimate SNR with good precision. However, since the influence of noise emphasis in ZF can be reduced, in general the characteristic is superior to that of ZF.
With the MLD algorithm, a transmit data stream is estimated according to the following equation:
      S    ^    =                    arg        ⁢                              min            k                    ⁢                                                                                      X                  -                                      HS                    k                                                                              2                        ⁢                                                  ⁢                          S              k                                          ∈                        {                                    S              I                        ⁢                          ⋯S              K                                }                ⁢                                  ⁢        K              =          Q      M      where Q represents the number of signal-point placements of the modulated data. In QPSK, Q=4; in 16QAM, Q=16; in 64QAM, Q=64. Thus, with MLD, the amount of calculation involved in multivalued modulation becomes very large, and the amount of calculation increases exponentially with respect to the number of transmit antennas. Since calculation of a matrix that is the inverse of a channel correlation matrix is made unnecessary by MLD, the relation N≧M is unnecessary.
With regard to the BLAST algorithm, the details are set forth in Non-Patent References 1, 2 mentioned later.
Generally, in a MIMO system, a transmission error tends to occur in a data stream that has been transmitted from an antenna for which the state of propagation path is inferior to other transmit antennas. Since the antenna for which the state of the propagation path is poor changes owing to fading fluctuation, the data stream that gives rise to the transmission error also changes with time. Further, in a MIMO system, transmission error tends to occur also in a case where the correlation between antennas increases owing to the propagation environment. More specifically, in a case where a path of particularly high power exists, such as a direct wave or strong reflected wave, in a multipath propagation path, the correlation between antennas increases. Since the state of the propagation path changes in a complex manner, there are instances where a transmission error tends to occur in a specific data stream owing to an increase in correlation between specific antennas. The state of such antenna correlation also changes from moment to moment owing to movement of a mobile station or a change in the surrounding environment.
Thus, with a MIMO system, there is a tendency for error to concentrate in a certain specific data stream, and the data stream in which error concentrates changes with time. In high-speed wireless data transmission, application of re-transmission control such as ARQ (Automatic Repeat reQuest) in radio intervals is essential. FIG. 17 illustrates a conventional example of a case where re-transmission control is applied in a MIMO system. Components identical with those shown in FIG. 14 are Designated by like reference characters. A data stream that has been multiplexed in space is demultiplexed by signal processing in the data processing unit DPU, demodulation/decoding processing is applied by a data demodulator/decoder RDU and the result is input to an error detector EDT. The latter performs error detection for every data stream ŝ1, ŝ2, . . . ŝM, and an ACK/NACK generator ANG reports the result (ACK/NACK) of error detection on a per-data-stream basis to the transmit station TR using the oppositely directed radio link (transmitter TX, transmit antenna ATT, receive antenna ATR, receiver RX). A re-transmission controller RTC of the transmit station TR performs re-transmission of a data stream applicable to NACK from among managing re-transmit buffers RTB1 to RTBM. The antenna used in transmission at this time is fixed. In other words, retransmission is performed using the same antenna as that used in the previous transmission.
In a case where re-transmission control is performed in a MIMO system, the antenna having the poor transmission path is used continuously when a re-transmission packet is transmitted from the same antenna. Consequently, a problem which arises is that the improvement in error rate by re-transmission is diminished and it becomes difficult to obtain re-transmission control gain. The problem becomes particularly acute if the change in fading or multipath environment is slow in comparison with re-transmission interval (round-trip time).
Patent Reference 1 is a first example of prior art of a MIMO system. In this first prior art, a receiver in a MIMO system detects the rate and power of each data stream and feeds these back to a transmitter, and the transmitter improves throughput by controlling the rate and power of the corresponding data stream based upon the rate and power that has been fed back. However, the first prior art does not improve error rate by re-transmission.
Patent Reference 2 is a second example of prior art of a MIMO system. In the second prior art, □ a transmitter creates at least two error-encoded streams from an information block and transmits these streams, and a receiver performs an error check on a per-stream basis and, if an error is detected, requests re-transmission only of an error-encoded stream for which an error has been detected, or □ a transmitter creates at least two error-encoded streams from an information block and transmits these streams, and a receiver combines the error-encoded streams, performs an error check and, if an error is detected, requests re-transmission of each error-encoded stream. Although the second prior art relates to re-transmission control, it does not improve error rate by re-transmission and does not raise re-transmission efficiency.